> Unfortunately that's the story for any OS these days that isn't already firmly established.
Maybe because we're coming at this from the wrong perspective?
I love the theoretical idea that I could build a generic x86 box that can boot into any OS I feel like using, but has that ever truly been the case? We certainly don't pick software this way—if you're running Linux, you're not going to buy a copy of Final Cut and expect it to work.
Well-established software will of course work almost everywhere, but niche projects don't have the ability. Unless you use something based on Java or Electron, which is equivalent to using Virtualbox (or ESXi) in this comparison.
It's long been said that one of Apple's major advantages with macOS is they don't need to support any hardware under the sun. Non-coincidentally, the recommended way to make a Hackintosh is to custom build a PC and explicitly select Mac-compatible hardware.
Now, if an OS doesn't for instance have support for any model GPUs at all, cherry picking hardware won't help. But perhaps this is where projects like BeOS need to focus their resources.
> The "correct" way to go about things is to choose the OS first, and then select compatible hardware.
Yeah, wouldn't it be nice if we weren't constrained by real world requirements? If I were to write an OS today, the hardware I'm targeting may become quite rare and/or expensive tomorrow. Or it may just go out of fashion. Regardless, very few people are going to buy new hardware just to try out an OS they're not even sure they want to use yet.
> very few people are going to buy new hardware just to try out an OS
We do have VM's and emulators, but yes, the cost of switching OS's is huge. That's true with or without broad hardware compatibility.
My point is this: I don't think the idea of OS-agnostic hardware ever really existed. The fact that most Windows PC's can also run Linux is an exceptional accomplishment, and not something other projects can be expected to replicate. You might get other OS's to boot, but not with full functionality.
Maybe because we're coming at this from the wrong perspective?
I love the theoretical idea that I could build a generic x86 box that can boot into any OS I feel like using, but has that ever truly been the case? We certainly don't pick software this way—if you're running Linux, you're not going to buy a copy of Final Cut and expect it to work.
Well-established software will of course work almost everywhere, but niche projects don't have the ability. Unless you use something based on Java or Electron, which is equivalent to using Virtualbox (or ESXi) in this comparison.
It's long been said that one of Apple's major advantages with macOS is they don't need to support any hardware under the sun. Non-coincidentally, the recommended way to make a Hackintosh is to custom build a PC and explicitly select Mac-compatible hardware.
Now, if an OS doesn't for instance have support for any model GPUs at all, cherry picking hardware won't help. But perhaps this is where projects like BeOS need to focus their resources.